Stochastic Duty Cycling for Heterogenous Energy Harvesting Networks

TL;DR

This paper introduces stochastic duty cycling algorithms for energy harvesting networks with heterogeneous and unpredictable energy sources, aiming to maximize active time while addressing energy inefficiency and variability.

Contribution

It proposes both offline and online duty cycling algorithms tailored for energy harvesting networks, with the online algorithm achieving a provable approximation ratio.

Findings

The online algorithm performs close to the offline optimal in experiments.

The online algorithm guarantees an approximation ratio of at least 1 - e^(-γ^2).

Experimental results validate the effectiveness of the proposed algorithms.

Abstract

In recent years, there have been several kinds of energy harvesting networks containing some tiny devices, such as ambient backscatter, ring and renewable sensor networks. During energy harvesting, such networks suffer from the energy heterogeneity, dynamics and prediction hardness because the access to natural resources is often spatiotemporal different and timely changing among the devices. Meanwhile, the charging efficiency is quite low especially when the power of the harvested energy is weak. It results in the energy waste to store the harvested energy indirectly. These features bring challenging and interesting issues on efficient allocation of the harvested energy. This paper studies the \emph{stochastic duty cycling} by considering these features with the objective characterized by maximizing the common active time. We consider two cases: offline and online stochastic duty cycling. For the offline case, we design an optimal solution: offline duty cycling algorithm. For the online case, we design an online duty cycling algorithm, which achieves the approximation ratio with at least $1-e^{-\gamma^2}$, where $\gamma$ is the probability able to harvest energy. We also evaluate our algorithms with the experiment on a real energy harvesting network. The experiment results show that the performance of the online algorithm can be very close to the offline algorithm.